DigiNews

Tech Watch by Johan Denoyer

← Back to articles

social media rabbit holes, clusters, and the relative mixing times of random walks

Quality: 8/10 Relevance: 9/10

Summary

This post analyzes a year-long Twitter data sample (2012) to identify affinity groups by extracting domain names from tweet links and visualizing them with PCA and t-SNE. It reveals tightly clustered right-wing domains compared to looser left-wing clusters and discusses how this affects recommender systems and random-walk mixing times on social feeds.

🚀 Service construit par Johan Denoyer